There are quite a few changes to the CMake build system packaged up into
one set here:
* Added `toolsrc/cmake/utilities.cmake`, which contains the following:
* `vcpkg_detect_compiler` -- get the name of the C++ compiler, as one
of {gcc, clang, msvc}
* `vcpkg_detect_standard_library` -- get the name of the standard
library we're linking to, as one of {libstdc++, libc++, msvc-stl}
* `vcpkg_detect_std_filesystem` -- figure out how to link and call
into C++17's filesystem; whether one needs to link to `stdc++fs` or
`c++fs`, and whether to use `<filesystem>` or
`<experimental/filesystem>`.
* Added a `VCPKG_WARNINGS_AS_ERRORS`, split off from
`VCPKG_DEVELOPMENT_WARNINGS`, which allows one to use the development
warnings without passing -Werror
* Rename `DEFINE_DISABLE_METRICS` to `VCPKG_DISABLE_METRICS` -- the
former will now print a deprecation message and set the latter.
* Now, print a deprecation message on `WERROR`; it doesn't do anything
since the behavior it requested is now the default.
* Pass `-std=c++17` if the compiler allows it, instead of `-std=c++1z`
* Do some code movement
* Pass `USE_STD_FILESYSTEM` if possible, instead of only on minGW
* Renamed to `VCPKG_USE_STD_FILESYSTEM`
Additionally, we now pass `/W4` in Debug mode on x86 in the Visual
Studio build system; this brings it in line with the CMake build system,
and the x64 Visual Studio build system.
And finally, we make some minor code changes to support compiling in
VCPKG_DEVELOPMENT_WARNINGS mode.
* [boringssl] Add new port
* removed usage of deprecated vcpkg functions
* let the ci treat boringssl like libressl
* applied requested changes
* [boringssl] Update to commit 590265773@2020-04-07 (#8455)
* [boringssl] Restore scripts/ci.baseline.txt
* [qt5-imageformats] freeglut is not a dependency on osx
* [qt5-base] fontconfig can require iconv on linux and osx
* include the jasper fix to get CI coverage?
* trying to get ci coverage for qt on osx
* Revert "trying to get ci coverage for qt on osx"
This reverts commit 8b9b188839.
* apply the correct ci hack
* fix wrapper to include macosx debug suffix
* update baseline
* first try to fix the Qt5 cmake configs at the root.
* fix missing backslash
* fix plugin patch
* fix qt config patch
* remove 90% of fixcmake.py
* update baseline
* experimental removal of fixcmake.py and dll copying into tools/qt5/bin
* fix experimental dll copying.
* fix static windows builds
* fix static builds
* fix plugin copy
* add quotes around configuration to make it work
* remove dll debug messages
* note about the cmake fix
* revert fontconfig changes on linux. Make qt.conf relative and not absolute
* make qt.conf relative
* trying to fix ogre osx ci regression
* fix silly path error
* Revert "trying to fix ogre osx ci regression"
This reverts commit 708d8250d6.
* add ogre:x64-osx=fail to CI baseline
* retry on osx
* fix configure error. since in qt5-base everything is in the package directory and not in the installed
* revert change to ci.baseline
* update ci baseline
* bump control
* update baseline
* ws change to osg-qt
* reset ci baseline to upstream/master
* update baseline
* fix osg-qt?
* skip ms-angle on osx and linux since the normal angle port should be used!
* update ci baseline
* try adding fontconfig manually
* update ci baseline to remove osg-qt for the time being
* remove manual library.
* New policy: SKIP_DLL_ARCHITECTURE_CHECK.
The check only works if MS link.exe is used
otherwise the second linker member is missing
(according to the observed errors)
* rename to VCPKG_POLICY_SKIP_ARCHITECTURE_CHECK
since the dll check was ok but the lib check was failing.
* fix indentation
* move the if to a better position.
* [vcpkg] Correct UInt128 code 😇
`UInt128::operator<<(x, y)` should clear the bottom 64 bits of `x` if
`y >= 64`; however, we don't do this, and so we duplicate `x`'s bottom
bits into `x.top` instead of moving them. Similarly, we have the
opposite problem for `UInt128::operator>>`. This commit fixes these
latent bugs, which we weren't hitting because the thing we use them for
never actually shifts more than 64 bits.
* add mirror for mpfr at gnu.org
* remove deprecated include(vcpkg_common_functions)
* update version number after modified port
* update copyright handling command to install
* fix configure meson for non windows builds.
* change to the two ports using meson
* fix passing of empty compiler arguments.
* dont transform / into - for all flags
* copy meson logs into the buildtree to better diagnose CI errors of meson ports
* another change to the two ports to actually force a CI rebuild
* add missing /pkgconfig dir to PKG_CONFIG_PATH
* fix pkgconfig setup
* add share/pkgconfig to PKG_CONFIG_PATH
* Add CMake path to PATH for meson
* add share folder to prefix for meson
* remove double message
* add libdir so that meson does not install into architecture folder in linux
and make prefix path target dependent. (Or should it be host dependent?)
* bump controls
* update toolchain selector to use VCPKG_TARGET_IS_<paltform>.
* change pkgconfig prefix and path.....
meson is not correctly handling native windows paths and eating "\"
PKG_CONFIG_PATH also nees : to correctly separate paths
Prefix still must be near to native path but with / instead of \ C:/somepath/
* added android triplets
* added android support to vcpkg
* added export directories to git ignore
* fix libraries naming
* added vckpg sources to visual studio project files
* rename file location
* issue with std::string fs:path copy initialization
* format path on VStudio
* fix checks format cannot work on fs::path
* support header only libraries
* support using architecture instead of triplets
* added prefab support
* added debug logs and prefab debug flag
* added support for empty packages i.e openssl
* [libffi] Support arm/arm64
* [libffi] Use VCPKG_TARGET_ARCHITECTURE to determine triplet
* [libffi] Use VCPKG_TARGET_ARCHITECTURE in CMakeLists.txt
* update baseline
* re-trigger ci test